Maaike H.T. de Boer, Romy A.N. van Drie, Roos M. Bakker and Daan Vos
It is often difficult to properly interpret legal texts. Even professionals do not always agree on the proper interpretation. It is, therefore, important to formalise the legal texts to make the interpretation explicit. One of the formalisation languages is named Flint (van Doesburg, 2019). The Flint language offers a formal model that enables the interpretation of legal text by describing the norms in these texts as acts and facts. Getting from legal texts to this formalisation requires a lot of time and effort from legal experts. In our research, we focus on the automatic extraction of Flint acts, starting with the components actor, action, recipient and object. These components have a resemblance with linguistic semantic roles. We created a dataset containing thousands of Dutch law sentences, manually annotated with the components of a Flint act, and applied a transformer-based method, a rule-based method and a hybrid method that combines both. We are currently analysing the results and we will share them at the conference. In future work, we plan to expand our automatic extraction with Flint facts, but also components such as pre-conditions.
References:
van Doesburg, R., & van Engers, T. M. (2019). Explicit Interpretation of the Dutch Aliens Act. In AIAS@ ICAIL (pp. 27-37).